Time-lapse imaging of subsurface flow patterns by ground penetrating radar

Abstract:

Introduction

The key to understand hydrological processes is the ability to observe and visualize them in natural environments. In hydrology, most processes such as subsurface flow can be considered to be preferential (Uhlenbrook, 2006). Consequently, our observation techniques require an adequate spatial and temporal resolution to accurately image the spatial and temporal characteristics of a specific preferential process. Considering the plot to hillslope scale, our process understanding is often based on dye tracer experiments or point measurements such as time domain reflectometry (TDR). Using dye tracers, a detailed spatial characterization of subsurface flow patterns is only possible after excavating the field site. However, the invasive nature of these experiments obviously limits the achievable repeatability and temporal resolution. In contrary, point based measurements are usually less invasive and provide a high temporal resolution. However, these methods lack the spatial resolution required to image and understand processes such as preferential flow. Consequently, there is a growing interest in using noninvasive geophysical methods, which promise a high spatial and a high temporal resolution for imaging subsurface architecture, parameters, and processes (Hyndman and Tronicke, 2005). Regarding such applications, a popular geophysical tool is ground penetrating radar (GPR). Under favorable site conditions, this technique provides the highest spatial resolution of any near-surface geophysical technique (e.g., Bristow and Jol, 2003; Jol 2009). In principle, GPR reflection profiling relies on transmitting an electromagnetic pulse from an antenna into the subsurface. At electromagnetic boundaries, defined by differences in the electromagnetic properties of subsurface media (electric permittivity and conductivity), the transmitted pulse is backscattered and recorded using a receiver antenna at the surface. After data processing, this backscattered energy can be interpreted in terms of a structural model, outlining subsurface architecture such as soil or geological layering. More challenging is the interpretation of GPR data with respect to physical parameters and parameter changes, respectively. Especially, in time-lapse GPR reflection imaging (e.g. employed to image subsurface flow paths) the applicability of the technique has been mostly demonstrated by controlled laboratory experiments (e.g., Pettinelli et al., 2013; Birken and Versteeg, 2000).

In this study, we evaluate the potential of time-lapse GPR reflection profiling to image subsurface preferential flow associated with an artificial infiltration experiment at a hillslope in Luxembourg. We demonstrate that after careful data acquisition and processing GPR is a capable tool to investigate subsurface flow patterns at our field site. By calculating difference images between 2D data sets recorded at different times along the same profile, we are able to highlight distinct subsurface areas associated with changes in soil moisture and interpreted as preferential flow pathways. Our results provide valuable information on the functional characteristics of preferential flow and may be used to evaluate functional and structural similarity as needed to develop a detailed understanding of the relevant hydrological processes at our field site. Experimental Site and Setup

We performed our infiltration experiment at a tributary to the Colpach River, which belongs to the Attert River basin located in the north-west of the Grand Duchy of Luxembourg. This site is underlain by Devonian schists, which are considered to be relatively impermeable (Bos, 2006). In addition, very rare observations of surface runoff have been reported, which indicates an important through-flow rate in the weathered bedrock (Krein et al., 2007). Further subsurface information originates from outcrop analyses by Juillerert (2011), who identified two distinct soil horizons. The upper layer is characterized by silty, loamy sediments with downwards increasing rock fragments whereas the basal layer consists mainly of angular, platy, and slightly weathered rock fragments. At our field site, this general soil structure is largely confirmed by geophysical investigations and coring data (not shown here). Furthermore, we observed air filled macropores in the basal layer. If connected, such macropores form ideal pathways and can contribute significantly to subsurface runoff processes. Thus, at our field site we expect strong lateral flow components because the bedrock is considered to be impermeable and the characteristics of the basal layer promote the formation of preferential flow paths.

Our experimental site covers an area of ~25 m x ~15 m on a forested north-facing hillslope with an average topographic gradient of ~14°. In the upper/southern part of our experimental site, four circular sprinklers have been installed in a square array (edge length 5 m x 5 m). This array provides a uniform rainfall distribution in the central area (5 m x 5 m) with steadily decreasing intensities outside. In the central area, an average intensity of 30 mm/h was realized for 4 hours. Compared to an annual rainfall of 852 mm (as reported by Pfister, 2000), our artificial precipitation experiment corresponds to an extreme rainfall event lasting for an unusual long time. In downhill direction, all precipitation below the central area was collected by a rain shield and drained to the outsides of our experimental site. Thereby, we ensured no direct precipitation on our four time-lapse GPR profiles, which were located at distances of ~0.5 m, ~1 m, ~2.5 m, and ~5 m downhill from the central infiltration area.

From initial experiments (not shown here), we learned that data quality and reproducibility is a critical issue when recording time-lapse GPR data. Thus, to ensure an accurate relocation of the measurement positions during time-lapse surveying, we fix the profile locations by wooden slats guiding the GPR antennas. In total, we collected GPR data at all four profiles at nine different time steps using a pulseEKKO Pro system with shielded 250 MHz antennas. Real time positions were acquired by combining our GPR system with an auto-tracking total station (Böniger and Tronicke, 2010). Processing of our GPR data relied on a standard GPR processing flow including band-pass filtering, zero time correction, amplitude scaling (based on correcting spherical divergence and absorption), fk-filtering, and topographic migration using a constant velocity of 0.07 m/ns. It should be noted, that we used only data independent processing steps to ensure a proper time-lapse interpretation of our data (e.g., for comparing amplitudes and calculating difference images). In addition to GPR surveying, we installed eighteen TRIME-PICO^®IPH/T3 sensors across our experimental area to measure depth profiles of soil moisture variations at depth intervals of 0.1 m and at distinct time intervals. The maximal installation depth of these probes varied from 0.4 m to 1.7 m depending on local soil conditions.

Results / Discussion

In Figure 1, we show selected results of our time-lapse GPR experiment and compare them to soil moisture data recorded at one position along the profile. In Figure 1d and 1e, we present processed GPR data recorded at two selected time steps. These data have been recorded along the profile being closest to the central infiltration area. These reflection images reveal complex reflection patterns indicating a rather complex subsurface architecture without well-defined subsurface horizons. Major GPR reflections originate from two depth intervals. One observed at a depth of ~0.8 m (~25 ns two-way travel-time (TWT)) and a second one at ~1.7 m (~50 ns TWT). These depth intervals are in good agreement with our coring data (not presented here) and correspond to the basic soil horizons as described above. Nevertheless, the structural information provided by GPR reflection imaging is rather obscure because no explicitly traceable reflections are present. However, the data recorded at different time steps (Figure 1d & e) show no clearly visible differences, which indicates the excellent repeatability and quality of our data. To highlight time-lapse variations in our GPR data as needed for a hydrological interpretation, we calculate the difference between each GPR profile and the reference profile presented in Figure 1d. From these difference images, we calculate the envelope and smooth the results using a horizontal running average filter with a kernel size of 0.40 m (corresponding to the estimated resolution limit of our GPR data). In doing so, we obtain eight difference images, highlighting time-lapse difference between GPR reflection profiles.
In Figure 1a to 1c, we present three selected difference images recorded after (~18 h and ~5 h) and before (~6 h) the infiltration event. Figure 1a shows only minor variations and can be considered as the noise level of our time-lapse data set. In Figure 1b, distinct areas of increased difference values (clearly beyond the noise level) are evident. These areas are vertically separated into two to three layers, whereas along the profile we observe two main areas of increased difference values (from profile distances ~5 m to ~7 m and ~8.5 m to >10 m). By comparing the maximum changes inferred from our GPR data with the data recorded by the soil moisture probe at one selected location on our profile (Figure 1f), we observe a clear temporal correlation between these two datasets. Consequently, we infer that our GPR difference images are linked to subsurface soil moisture changes and areas of increased differences can be interpreted as preferential flow paths.
Surprisingly, the difference image recorded before the start of the infiltration (Figure 1c) does not reach the same low difference values considered as noise level in Figure 1a. Although the soil moisture data show no significant changes at this time step, there are other areas in this difference image (at profile distances of ~2 m to ~5 m and ~6.5 to 8) showing even higher difference values compared to the images shown in Figure 1a and 1b. We explain this observation with a natural rainfall event of ~10 mm/h (lasting for 3 h) occurring ~23 h before the start of the artificial precipitation event. Assuming that our difference images indicate the amount of water flowing in the subsurface, we observe a stronger flow from the natural event compared to the artificial one. Considering that natural precipitation affects the entire hillslope expanding ~300 m further upslope, we estimate ~3000 m³ of water (considering a 10 m wide slope section) draining through our GPR profile. Comparing this to the water infiltrated during our infiltration experiment (4 m3) provides a reasonable explanation for our time-lapse GPR data recorded before the start of our infiltration experiment.

Conclusions

Using data collected during an artificial infiltration experiment, we have evaluated the feasibility of time-lapse GPR reflection profiling to image subsurface flow phenomena. We calculated difference images of GPR data collected along the same profiles before, during, and after artificial precipitation. Using these difference images, we were able to identify distinct areas of increased difference values, which show a strong temporal correlation with soil moisture point measurements. As previous studies have largely focused on imaging processes under controlled laboratory conditions, our results clearly show the potential of time-lapse GPR surveying to image preferential flow paths under real field conditions. Because we are also able identify subsurface changes associated with a natural precipitation event, we believe that GPR has the potential to image subsurface flow paths from artificial and natural precipitation and infiltration events, respectively. From a methodological point of view, we found that a careful survey design ensuring high data quality and positioning repeatability is critical for such time-lapse reflection surveys. Similarly, careful processing strategies including an adequate migration scheme are needed to image time-lapse phenomena in a reliable fashion.

References

Birken, R. and Versteeg, R.: Use of four-dimensional ground penetrating radar and advanced visualization methods to determine subsurface fluid migration, Journal of Applied Geophysics, 43(2-4), 215–226, 2000.Böniger, U. and Tronicke, J.: On the potential of kinematic GPR surveying using a self-tracking total station: evaluating system crosstalk and latency, IEEE Transactions on Geoscience and Remote Sensing, 48(10), 3792–3798, 2010.

Bos, R. van den, Hoffmann, L., Juilleret, J., Matgen, P. and Pfister, L.: Conceptual modelling of individual HRU’s as a trade-off between bottom-up and top-down modelling, a case study, in Proceedings of the iEMSs third biennial meeting: summit on environmental modelling and software, 2006.

Bristow, C. S. and Jol, H. M.: Ground penetrating radar in sediments, The Geological Society of London, London, 2003.

Hyndman, D. and Tronicke, J.: Hydrogeophysical Case Studies at the Local Scale: The Saturated Zone, in Hydrogeophysics, vol. 50, edited by Y. Rubin and S. S. Hubbard, pp. 391–412, Springer Netherlands, Dordrecht, 2005.

Jol, H. M.: Ground penetrating radar theory and applications, 1st ed, Elsevier Science, Amsterdam, 2009.

Juilleret, J., Iffly, J. F., Pfister, L. and Hissler, C.: Remarkable Pleistocene periglacial slope deposits in Luxembourg (Oesling): pedological implication and geosite potential, Bulletin de la Société des Naturalistes Luxembourgeois, 112, 125–130, 2011.

Krein, A., Salvia-Castellvi, M., Iffly, J.-F., Barnich, F., Matgen, P., Bos, R. v.d., Hoffmann, L., Kies, A. and Pfister, L.: Uncertainty in chemical hydrograph separation, in Uncertainties in the “monitoring-conceptualisation- modelling” sequence of catchment research, edited by L. Hoffmann and L. Pfister, pp. 21–26, International Hydrological Programme (IHP) of the United Nations Educational, Scientific and Cultural Organization (UNESCO), Luxembourg, 2006.

Pettinelli, E., Di Matteo, A., Beaubien, S. E., Mattei, E., Lauro, S. E., Galli, A. and Vannaroni, G.: A controlled experiment to investigate the correlation between early-time signal attributes of ground-coupled radar and soil dielectric properties, Journal of Applied Geophysics, 101, 68–76, 2014.

Pfister, L., Humbert, J. and Hoffmann, L.: Recent trends in rainfall-runoff characteristics in the Alzette river basin, Luxembourg, Climatic Change, 45, 323–337, 2000.

Robinson, D. A., Binley, A., Crook, N., Ferr, T. P. A. and Grauch, V. J. S.: Advancing process-based watershed hydrological research using near-surface geophysics: a vision for, and review of, electrical and magnetic geophysical methods, Hydrological Processes, 22, 3604–3635, 2008.

Uhlenbrook, S.: Catchment hydrology - a science in which all processes are preferential, Hydrological Processes, 20, 3581–3585, 2006.